Changes in carbon stocks related to land use and land-use change

Transcription

1 Appendix C Changes in carbon stocks related to land use and land-use change 1. Introduction This appendix discusses GHG emissions and changes in carbon stocks that result from land use and land-use change. Land uses and LUCs are defined; the relevant carbon pools and emission sources are discussed in the context of these categories; the approaches to estimating emissions and changes in carbon stocks are outlined; and, finally, justification for, and an explanation of, the selected estimation methods used in this study are also provided. Land use, land-use change and forestry (LULUCF) is defined by the United Climate Change Secretariat as: a greenhouse gas inventory sector that covers emissions and removals of greenhouse gases resulting from direct human-induced land use, land-use change and forestry activities. Six land use categories are defined in the 2006 IPCC Guidelines for National Greenhouse Gas Inventories: 1. forest land; 2. cropland; 3. grassland; 4. wetlands; 5. settlements; and 6. other land. Land may remain in any of these categories or, in the case of LUC, its use may change to another category (e.g. from forest to grassland). Thus, each land use category can be further subdivided into land that is converted from one land use category to another, and land that remains in the same category. While this study focuses on the emissions from LUC, emissions from land use are also discussed. 1.1 GHG emissions from land-use change Most LUCs alter the soil and vegetation of the land, thus changing the amount of carbon stored per unit area. These changes may be positive or negative, and may occur in each carbon pool: biomass (above- and below-ground); dead organic matter (dead wood and litter); and soil (soil organic matter). LUC can significantly alter the carbon stored in biomass, by replacing the vegetation of the existing land use category with the vegetation of another land use category. Conversion of forest land to either grassland or cropland can lead to large and rapid losses of the typically large stores of carbon in forest vegetation, when this vegetation is replaced with herbaceous grasses or annual crops. While most of the carbon stored in forest biomass is lost following conversion, some carbon will be transferred from one pool to another; e.g. when trees are felled, a portion of the above-ground biomass is transferred to the dead organic matter pool, and a portion of the below-ground biomass is transferred to the soil organic matter pool. The drainage and cultivation or grazing of organic soils is also an important cause of the oxidation and loss of soil organic carbon (SOC) for both croplands and grasslands (Armentano and Menges, 1986). While the most important GHG emission flux is CO 2, the oxidization of the various organic carbon pools as a consequence of LUC can also release N 2 O. 123

2 Greenhouse gas emissions from pig and chicken supply chains Land conversion, often results in an abrupt change where most biomass is lost, followed by a longer period where biomass is oxidized at a much slower pace. The IPCC (2006) assumes a default 20-year transition period following conversion over which all losses are accounted for. The conversion of forest land to agricultural land may also lead to losses from the SOC pool. When forest land is converted to cropland, there is an average reduction in soil carbon of between 25 and 30 percent in the upper metre of soil (Houghton and Goodale, 2004). 13 These soil carbon losses are due, in part, to a lower fraction of non-soluble material in the more easily decomposed crop residues, and to the breaking up of aggregates and subsequent exposure of organo-mineral surfaces to decomposers following tillage (Kwon and Post, 2000). On the other hand, because grasslands, unlike crops, are not ploughed (temporary cultivated pastures are classified to be crops), little change in soil carbon is expected following the conversion of forests to grasslands (Houghton and Goodale, 2004). When either cropland or grasslands are abandoned, there is a re-accumulation of carbon in vegetation as the land returns to its natural state and the greater the biomass of the returning vegetation the larger is the long-term carbon sink due to the recovery. Post and Kwon (2000) note relatively low rates of accumulation in mineral soil following the abandonment of cropland. Considering all LUCs during the 1990s, Houghton and Goodale (2004) estimate that the average annual emissions from LUC is 2.2 petagram C/year, with almost all of this emanating from deforestation in the tropics. 1.2 Land use and its effects on emissions and carbon stocks Agricultural lands hold substantial carbon stocks, mostly in soil organic matter. Carbon stock changes in agricultural lands are closely tied to management practices, which can either enhance or erode carbon stocks. Practices that raise (lower) the photosynthetic input of carbon and/or slow (accelerate) the release of stored carbon through respiration, erosion or fire will increase (decrease) carbon stocks (Smith et al., 2007). While it is possible for substantial biomass carbon to be stored through perennial plantings on agricultural lands (e.g. silvopastoral systems), carbon accumulation and losses occur mostly in the SOC pool. This below-ground carbon pool also has slower rates of turnover than above-ground pools, because most of the organic carbon in soils comes from the conversion of plant litter into more persistent organic compounds (Jones and Donnelly, 2004). Smith et al. (2007) estimated that 89 percent of the agriculture sector s total mitigation potential is from SOC sequestration. For croplands, significant changes in SOC stocks are associated with management practices including tillage, residue management, nutrient management and the use of organic amendments (Smith et al., 2007). Historically, while agricultural management practices can result in either reductions or accumulations in the SOC pool, agricultural lands are estimated to have released more than 50 petagram C (Paustian et al., 1998; Lal, 1999, 2004a), some of which can be restored via better management. Currently, however, the net flux of CO 2 between the atmosphere and agricultural lands is estimated to be approximately balanced (Smith et al., 2007). For the estimation of net livestock sector GHG 13 While there is some variation around this range, it has been documented in numerous studies, and has been found to be broadly robust across all ecosystems (Houghton and Goodale, 2004). 124

3 Appendix C Changes in carbon stocks related to land use and land-use change emissions, which is the main purpose of this report, measures of net CO 2 current fluxes by region are of greater interest than the sequestration/mitigation potential. The lack of a globally consistent and regionally detailed set of net CO 2 flux estimates make it difficult to quantify these potential emission sources and sinks by region in this study, although there are some relevant studies that provide useful estimates of these net fluxes for specific regions and agricultural land use categories. For example, based on literature observations for temperate grasslands mainly from Western Europe, Soussana et al. (2010) estimate that grasslands SOC sequestration rates averaged 5 ± 30 g C/m 2 per year. There is also considerable potential to sequester carbon in croplands through a range of available options that include reduced and zero tillage, set-aside, perennial crops, deep-rooting crops, more efficient use of organic amendments, improved rotations, irrigation, etc. In Brazil, for example, long-term field experiments (Costa de Campos et al., 2011; Dieckow et al., 2010; Vieira et al., 2009; Sisti et al., 2004) have evaluated the impact of conservation tillage and crop rotations on SOC. The results from these studies confirm that non-tillage and crop rotations can promote the conservation of SOM and increase C accumulation. For example, Dieckow et al. (2010) assessed the 17-year contribution of no-tillage crop rotations to C accumulation in the subtropical Ferralsol of Brazil and concluded that crop-forage systems and crop-based systems with legume represent viable strategies to increase soil organic C stocks. They found that alfalfa systems with maize every three years showed the highest C accumulation (0.44 tonnes C/ha/yr). The biannual rotation of ryegrass (hay)-maize-ryegrass-soybean sequestered 0.32 tonnes C/ha/yr. However, an assessment of realistically achievable potentials for carbon sequestration in croplands needs to take into account economic, political and cultural constraints as well as other environmental impacts (such as non-co 2 GHG emissions). 2. Quantification of carbon emissions and SEQuestration 2.1 Changes of carbon stocks related to land-use changes The most fundamental step in assessing emissions from LUC is the tracking of changes in areas of land use and conversions from one land use category to the next. This tracking requires a time series of data, or data collected from at least two points in time, to capture changes in the area of land for each category. Comprehensive guidance on methodological approaches for estimating LUCs as well as emissions and removals from LULUCF is provided in the 2006 IPCC Guidelines for National Greenhouse Gas Inventories (IPCC, 2006). Three different approaches are suggested with differing degrees of accuracy, to best ensure the consistent representation of LUCs for given data quality and availability. The most accurate of these, Approach 3, requires the use of spatially-explicit data for land use categories and conversions, and includes the use of gridded map products derived from remote sensing imagery. At the other extreme is Approach 1, which relies on non-spatially explicit data from census and survey data, often reported at country or province level, and which only permits net changes in land use categories over time, and cannot specify inter-category conversions. Finally, Approach 2 enables the tracking of conversions between land use categories without the spatially-explicit location data. Naturally, the choice between the simple and the more sophisticated approaches involves big tradeoffs between the data and analytical resource requirements, and the accuracy with which LUCs and their attendant emissions and carbon removals are estimated. 125

4 Greenhouse gas emissions from pig and chicken supply chains For grassland remaining grassland, cropland remaining cropland, and conversion from forestland to either of these land use categories, the 2006 IPCC Guidelines require that changes in carbon stocks from each carbon pool (i.e. above-ground biomass, below-ground biomass, dead wood, litter and soil organic matter), as well as emissions of non-co 2 gases, are estimated. The guidelines do, however, provide flexibility in the use of methods that range from very simple approaches that rely on default EFs to more sophisticated approaches that use detailed location-specific data and process models that fully characterize the fluxes between carbon pools Biomass and dead organic matter (DOM) pools As mentioned, land-use conversions are often associated with an initial abrupt change and subsequent transition period following conversion. The 2006 IPCC Guidelines provide separate equations for these two phases when using Tier 2 and 3 approaches. Where country-specific EFs are available and comprehensive national data are available, country-defined Tier 3 methodologies based on either process models or detailed inventories, stratified by climate and management regime can be recommended. These methods can also use non-linear loss and accumulation response curves during the transition phase. At the other extreme, Tier 1 methods assume that both biomass and DOM pools are lost immediately after conversion from forestland to agricultural land, and that agricultural land reaches its steady-state equilibrium in the first year following conversion. While the IPCC provide default values to quantify biomass levels prior to and after conversion, there is assumed to be no accumulation in the DOM pool in the transition phase on agricultural land following conversion from forestland. The Tier 2 methods represent a compromise, better capturing the dynamics of land-use conversion, by specifying separate equations for the abrupt change and transition phases, accounting for biomass accumulation during the latter phase. They also rely on some country-specific estimates of initial and final biomass stocks, instead of relying solely on default values. Further, both Tier 2 and 3 methods account for transfers between carbon pools and can estimate carbon pool changes using either the gain-loss or stock-difference methods. The former method includes all processes that cause changes in a carbon pool, including biomass growth and the transfer of carbon from one pool to another. Alternatively, the stock-difference method can be used where carbon stocks are measured at two points in time. Both methods are valid, providing they can represent disturbances and continuously varying trends, and can be verified with actual measurements (IPCC, 2006) Soil organic carbon (SOC) stocks Changes in the SOC pools in both mineral and organic soils should be taken into account when estimating emissions and carbon accumulation resulting from LUC (IPCC 2006). In order to account for these changes, areas of converted land must be stratified by climate region, management and major soil type. Simple Tier 1 methods, which rely on default reference SOC stock change factors, can be used, or more country- or region-specific reference C stocks and stock change factors can be combined with more disaggregated land use activity data to use either Tier 2 or Tier 3 methods. Some of the process models suited to Tier 3 methods are discussed in the following section. 126

5 Appendix C Changes in carbon stocks related to land use and land-use change In this study, the LUC emissions for the major carbon pools, i.e. biomass, DOM and SOC pools, are estimated using Tier 1 methods. While Tier 2 and Tier 3 methods are recommended, the Tier 1 approach was deemed to be appropriate given the global nature of the assessment combined with the absence of country-specific EFs, inventory data and/or a suitable global process model. 2.2 Changes in carbon stocks for agricultural land remaining in the same land use category As with LUCs, the estimation of emissions and carbon accumulation from management practices on land that remains in the same land use category requires that changes in carbon stocks from each major carbon pool (i.e., above-ground biomass, below-ground biomass, dead wood, litter and soil organic matter), as well as emissions of non-co 2 gases, are estimated. For agricultural lands, changes in these carbon pools and non-co 2 emission fluxes depend on management practices such as grazing, burning, pasture management, tillage and residue management. Tier 2 and Tier 3 methods are able to estimate changes in each carbon pool and in emissions resulting from management practices, while Tier 1 methods can only be used to estimate these changes for the SOC pool (and non-co 2 emissions from burning), but not for the other carbon pools. As with the measurement of emissions and carbon storage under LUC, the same gainloss and/or stock-difference methods can be employed for land use estimates. As discussed, Tier 3 methods can be used to more accurately assess changes in these carbon pools and non-co 2 emission sources, using dynamic process models and/or detailed inventory measurements to estimate carbon stock changes. Process model-based approaches simultaneously solve multiple equations to estimate net changes in carbon stocks. These models can incorporate management effects such as grazing intensity, fire, fertilization, tillage and residue management, and they can be combined with regionally representative sampling-based estimates to validate and extrapolate to other agricultural lands. According to IPCC (2006), important criteria for selecting these models include: their ability to represent all relevant management practices and production systems, the compatibility of model s driving variables (inputs) with available country data, and validity gauged by the model s ability to represent stock change dynamics reported in empirical assessments. Well-known biogeochemical models that can satisfy these criteria include the Century model (and the daily time-step version, Daycent), DNDC and RothC. The RothC (Hart, 1984; Jenkinson et al., 1987; Coleman et al., 1997; Smith et al., 2006) and Century (Parton et al., 1987; Falloon and Smith, 2002; Kirschbaum and Paul, 2002) models can be used to simulate GHG gas exchange and carbon cycling dynamics of cropland, grassland and forestland land use categories, and both operate on monthly time-steps. Soil texture and weather data are the major input variables. While the Century model can simulate the dynamics of carbon in biomass, DOM and SOC pools, as well as nitrogen, phosphorous, and sulphur dynamics, RothC only estimates SOC stocks and CO 2 losses from decomposition of SOC. The Daycent model is the daily time-step version of the Century model (Del Grosso et al., 2001; Parton et al., 1998), which is well suited to capturing N mineralization and N gas production in non-waterlogged soils, along with the same carbon pool dynamics modelled in Century. As with Daycent, the denitrification-decomposition (DNDC) model (Li, 1996; Li et al., 1992, 1994) simulates soil carbon and nitrogen 127

6 Greenhouse gas emissions from pig and chicken supply chains fluxes using a daily time-step but, unlike Daycent, it is also able to represent N gas and CH 4 fluxes from waterlogged soils, such as are found in rice paddies. Both Daycent and DNDC have higher data demands than either Century or RothC, due their short time-steps and wider range of biogeochemical dynamics. Since none of these models have been validated on a global scale, they have not been applied in this analysis. 3. Quantification of carbon STOCk changes from land use and land-use change in this report In this study, LUC emissions are estimated for three major carbon pools, including the biomass, DOM and SOC pools. It could be argued that Tier 2 and Tier 3 methods, including process-based modelling approaches, should have been used to capture variability and possibly to reduce uncertainty in the emission and carbon accumulation estimates. However, given the global nature of the assessment, and the absence of country-specific EFs, carbon stock/flux inventory data and/or a suitable global process model (cf. previous section), the Tier 1 approach was deemed a suitable option to develop preliminary estimates and shed light on the potential magnitude of the LUC emissions for the sector. For the reasons outlined above, this assessment does not cover changes in C stocks occurring under constant land use management. This may be done in future updates once global datasets are available and/or models have been calibrated for global studies. This section presents the approach applied in this study to quantify LUC emissions, discussing the rationale for the approach chosen, and the results from the analysis. It also explores the implications of alternative approaches to quantifying LUC emission. 3.1 Approach The analysis focuses on one specific feed product soybean in specific countries in Latin America. This assessment is based on observed land use trends, feed crop expansion trends and trade flow patterns as well as findings from previous studies such as Wassenaar et al. (2007) and Cederberg et al. (2011). This study uses IPCC guidelines as a basis for the quantification of LUC emissions. This choice is largely based on the fact that the IPCC approach meets the UN- FCCC needs for calculating and reporting of GHG emissions from LUC. The cropland part of this assessment also relies on other guidelines such as the PAS 2050 (also based on IPCC guidelines) for input data. According to IPCC Guidelines, emissions arising from LUC are allocated over a 20-year period (the amortization period). Because of data availability (forestry inventories are only available from ), in this assessment, the rates of LUC are taken as the average over the 16-year period ( ). This practically discounts four years of emissions. Agriculture has been a major driving force behind land transformation; globally, the area of land used for agriculture increased by 83 million ha over the period In most regions, cropland has increased whereas pasture and forest land decreased (Figure C1). The most affected regions in terms of crop expansion are Latin America, Asia and Africa. Declining agricultural land (i.e. cropland and pastureland) is observable in Europe and North America where agricultural land abandonment 14 The FAOSTAT forest area dataset (based on the Global Forest Resource Assessment) used in this study is only available from 1990 and in order to align the C stocks assessment with the livestock input data which is based on 2005 statistics, land-use conversion trends were assessed for the period 1990 to

7 Appendix C Changes in carbon stocks related to land use and land-use change Figure C1. Net land conversion between 1990 and 2006, by region Cropland Pastureland Forest Other land Oceania LAC N. America Europe Asia Africa , hectares Source: FAOSTAT (2012). Table C1. Global area expansion for selected crops with highest area expansion ( ) Crop Area expansion (1 000 ha) Share of global gross crop expansion (percentage) Soybeans Maize Rapeseed Rice, paddy Sunflower seed Oil palm fruit Source: FAOSTAT (2012). has resulted in reforestation. During the period considered ( ), deforestation occurred mainly in Africa and Latin America. More recent trends in deforestation, in particular in Latin America (reduced deforestation rates) and Asia (increasing rates), and their association with feed production are therefore not considered in this study. Between 1990 and 2006, crop expansion was mainly driven by major oil crops (e.g. soybeans, rapeseed, sunflower and oil palm) the demand for which was fuelled by demand for vegetable oil, feed and, latterly, biofuel policies. The expansion of soybean production is argued to be one of the major drivers of LUC, particularly deforestation (Pacheco, 2012; Nepstad et al., 2006; Fearnside, 2005; Bickel and Dros, 2003; Carvalho et al., 2002). The global area under cultivation of soybean has increased rapidly in recent decades; between 1990 and 2006, the global soybean area increased faster than any other crop (Table C1). Maize expansion is also important, representing 9.2 percent of global crop expansion. At the same time, some other crops, such as wheat, barley, and oats, have strongly declined, which explains the apparent discrepancies with the net land conversion trends in Figure C1. 129

8 Greenhouse gas emissions from pig and chicken supply chains Figure C2. Maize and soybean area expansion between 1990 and 2006, by region 30 Maize Soybean 25 Million hectares Africa Asia Europe N. America LAC Source: FAOSTAT (2012). A comparison of the two major crops driving agricultural expansion reveals key regional differences with regard to their importance (Figure C2). The expansion of soybean area has been significant in North and South America, while maize expansion is more important in Africa and Asia. Deforestation for crop expansion has been an important LUC process in Africa, however crop expansion in the region has been mainly driven by sorghum and millet, with maize and soybeans only accounting for 5 percent and 0.5 percent of total gross cropland expansion respectively. In Africa, pasture expansion has also occurred largely at the expense of forest area. However, due to lack of reliable data and information it is difficult to draw conclusions on the land-use conversion trends in this region. In North America, soybean expansion is responsible for 37 percent of total crop expansion and maize 7 percent. However in this region the overall trend has been a decrease of total cropland (due to sharp decreases in wheat and barley areas) and pastures and an increase of forest area. In Asia, soybean expansion is responsible for 7 percent of total crop expansion and maize 8 percent. At the same time, forest land has increased overall in Asia and pastureland has decreased. But the two trends occurred in different subregions within Asia. Pasture decrease mainly occurred in Mongolia and Iran, where maize and soybean expansion were null or limited. On the contrary, expansion of soybean and maize area has largely occurred in India and China (77 percent of gross maize expansion and 96 percent of gross soybean expansion), but forest area increased in these two countries. Pastures decreased in India but to a limited extent of 1.2 million ha, compared to the 5.8 and 3.0 million ha of, respectively, soybean and maize expansion in the country. In Latin America, most of the decrease in forest area happened in the countries where soybean expansion was occurring. Trends in land conversion, particularly deforestation, are therefore closely linked to the expansion of soybean. 130

9 Appendix C Changes in carbon stocks related to land use and land-use change Table C2. Average annual land-use change rates in Argentina and Brazil ( ) Land-use type Argentina Brazil (1 000 ha) Agricultural area Grasslands Arable land & permanent crops Soybean area Forest area Other land Source: FAOSTAT (2009). Based on these observations the scope of our assessment was reduced to the soybean expansion in Latin America. Within Latin America, Brazil and Argentina account for 91 percent of the total soybean area. In the period , 90 percent of the soybean area expansion in Latin America took place there, further narrowing the scope to these two countries. An assessment of land use trends in the key producing regions shows that the expansion in soybean area has been largely gained at the expense of forest area (Table C2). In Argentina, the annual increase of area dedicated to soybean is much larger than the increase of total arable land (Table C2), indicating that there has been a shift in land use from other crops to soybean. According to FAOSTAT statistics, 44 percent of the new soybean area was gained against other crops, while the rest was gained against forest (22 percent) and other land (31 percent). The latter category covers natural vegetation that does not include forest and grazed natural grasslands. The reported annual increase of soybean area in Brazil is ha (Table C2). We assumed a simplified pattern of deforestation in the Amazon, in which cleared land is first used as pasture and/or crop land, and then possibly left as fallow land. The latter, classified as other land in FAOSTAT, is occupied by weeds, grasses, shrubs and, partly, by secondary forest. Under this assumption, every year roughly 2.9 million ha are converted to arable land and grassland during the period covered in this assessment. At the same time, agricultural land is abandoned at a rate of 1.6 million ha per year. The annual net increase of arable land and grassland is 0.53 and 0.75 million ha, respectively. We thus assume that all incremental soybean area is gained at the expense of forest area. Rates of C loss/gain arising from specific land-use transitions were taken from PAS 2050 guidelines (BSI, 2008), which are also based on IPCC (2006). These estimate deforestation (conversion of forest to annual cropland) releases in Brazil at an average 37 tonnes CO 2 -eq/ha, and conversion of forest and shrub land to annual crop in Argentina at 17 and 2.2 tonnes CO 2 -eq per ha, respectively. GHG emissions from soybean-driven LUC were calculated as the accumulated emissions for one year resulting from the total area deforested during the period divided by the total soybean production in Based on this data, two LUC emission intensities were estimated for soybean cake produced in Brazil and Argentina, respectively: 7.69 and 0.93 kg CO 2 -eq/kg soybean cake. Soybeans and soybean cake produced elsewhere were assumed not to be associated with LUC. 131

10 Greenhouse gas emissions from pig and chicken supply chains Table C3. Regional sources of soybean and soybean cakes in 2005 (percentage) Brazil Argentina Other Soybean Soybean Cake Soybean Soybean Cake Soybean Soybean Cake LAC E & SE Asia E. Europe N. America Oceania Russian Fed South Asia SSA NENA W. Europe Source: FAOSTAT (2013). Table C4. Main exporters of soybean and soybean cakes in 2005 Exports (Million tonnes) Soybean Share of global exports (percentage) Exports (Million tonnes) Soybean cake Share of global exports (percentage) Argentina Brazil United States of America India Paraguay Source: FAOSTAT (2013). Table C5. Land-use change emissions associated with soybean production Region Pigs Chicken Cattle (Million tonnes CO 2 -eq) LAC East Asia E. Europe N. America Oceania Russian Fed South Asia SSA NENA W. Europe World Source: GLEAM. 132

11 Appendix C Changes in carbon stocks related to land use and land-use change In quantifying total emissions associated with the transformation of forest for soybean cultivation, LUC emissions are attributed to only those countries supplied by Brazil and Argentina with soybean and soybean cake. Table C3 gives the share of soybean and soybean cake sourced from Brazil and Argentina for each region, and Table C4 gives the main exporters. 3.2 Results for land-use change This analysis shows that about 224 million tonnes CO 2 -eq arise per annum from the expansion of soybean production in Brazil and Argentina to meet global demand for pigs, chickens and cattle feed. The bulk of these emissions arise in response to soybean consumption in Europe, East Asia and LAC (Table C5) which source large quantities of their soybean feed from Argentina and Brazil. The emissions estimated for the livestock sector in Western Europe are particularly high, which not only indicates a high reliance on imported soybean and soybean cake for feed, but also use of soybean with a high emission intensity, particularly because a large share is sourced from Brazil (see Table C3). Meeting demand for pig and chicken feed accounts for 195 million tonnes CO 2 - eq per annum, 87 percent of the total. This result is not surprising because of the high share of soybean in the diets of these species. Regarding the cattle sector, LUC emissions from soybean are important in Europe where it is utilized in dairy production. The results suggest that emissions are largely influenced by: (i) the quantity of soybeans and soybean cake imported from the two countries; and (ii) the share of soybean in the ration. The results presented here are part of the ongoing process of improving the estimation of LUC emissions. In order to progress towards better methodologies, certain gaps in data and in scientific understanding need to be addressed. The following section outlines some of the challenges and investigates the influence that methodological choice can have on LUC emissions. 3.3 Sensitivity analysis and the influence of land-use change method Modelling of land use and LUC emissions is subject to great uncertainties mainly because of the complexity of LULUCF processes, the challenges of obtaining reliable global data and the absence of validated approaches to estimate carbon stock changes. In particular, uncertainty regarding the magnitude of LUC emissions arises due to uncertainties in: (a) the rates of land use; (b) the carbon storage capacity of different forests, initial carbon stocks and the modes of C release; and (c) the dynamics of land use not normally tracked. In addition, a value judgment has to be made regarding what drives LUC and, consequently, how the emissions should be allocated. In order to explore the potential effect that different methodologies can have, the results obtained with the GLEAM approach are compared to three alternative approaches: (a) PAS :2012; (b) One-Soy; and (c) reduced time-frame approach. These approaches are summarized in Table C Alternative approaches PAS : 2012 approach. Several studies suggest that deforestation is related to the expanding soybean sector (Fearnside, 2005; Bickel and Dros 2003; Carvalho et al., 2002), but others dispute this claim, and argue that soybean is expanding into land previously under pasture, and is not causing new deforestation (Mueller, 2003; 133

12 Greenhouse gas emissions from pig and chicken supply chains Table C6. Alternative approaches for soybean LUC emissions calculations Method Spatial allocation Temporal allocation of LUC emissions (amortization) Quantification of rates of LUC GLEAM approach (current study) PAS :2012 One-Soy Reduced time-frame Source: Authors. To all soybean produced within the country To all soybean produced within the country To traded soybean To all soybean produced within the country 20 years FAOSTAT average LUC rates Brazil: forest crops (100%) Argentina: other crops (44%), forest (22%) and other land (31%) soybean 20 years Average rates over 20 years. LUC rates based on (a) or (b) - whichever results in the highest emission factor. (a) from grassland forest and perennial arable in equal proportion (b) from grassland, forest and perennial arable in proportion to their rates of change 20 years FAOSTAT average LUC rates Brazil: forest crops (100%) Argentina: other crops (44%), forest (22%) and other land (31%) soybean 20 years FAOSTAT average LUC rates Brazil: forest crops Argentina: other crops (44%), forest (22%) and other land (31%) soybean Quantification of rates of C loss/gain IPCC (2006) Tier 1 IPCC (2006) Tier 1 IPCC (2006) Tier 1 IPCC (2006) Tier 1 Brandao et al., 2005). Due to the lack of knowledge of the origin of the converted land, the GLEAM results were compared with PAS :2012 (BSI, 2012), which provides a way of quantifying LUC emissions when previous land use is not known and only the crop and country are known. The PAS :2012 calculations of emissions related to land-use change are accomplished in two steps. First, rates of land-use change need to be calculated based on the PAS : To calculate these, four categories of land are considered: forest, pasture, annual cropland and perennial cropland. Time series data on land area for forest, pasture, annual and perennial crops taken from FAOSTAT were used to: (i) determine whether the crop in question was associated with LUC by quantifying the rate of expansion over a 20-year period; and (ii) determine the share of LUC associated with each land category. In a second step, carbon losses based on land dynamics and biophysical conditions (climate, soil type, forest type, crop management, etc.) were computed based on the IPCC (2006) Tier 1 approach. The two sources of carbon taken into account in this approach are vegetation and soil. Two LUC EFs were calculated, based on different assumptions regarding where land for soybean expansion is derived from: (i) assuming that land for soybean production is gained in equal proportions from grassland, forest and perennial cropland; (ii) assuming that land for soybean is gained from other land use categories in proportion to their relative rates of change. The highest of the two EF s was then selected, in accordance with the guidelines. BSI (2012) present a detailed account of methodology and data sources. One-Soy approach. In this approach it is assumed that all soybeans, irrespective of where they have been produced, are associated with LUC. The central argument for this scenario is that the global demand for soybeans is largely interconnected and is a key driver of LUC. An average LUC emission factor associated with soybean was 134

13 Appendix C Changes in carbon stocks related to land use and land-use change Table C7. Summary of land-use change emission intensity in current study: alternative approaches for soybean cake Scenario Argentina Brazil (kg CO 2 -eq per kg soybean cake) GLEAM approach (current study) PAS : One-Soy Reduced time-frame Source: Authors calculations. estimated by calculating the total LUC emissions attributable to globally-traded soybean and soybean cake and then dividing this by total global soybean cake exports. Because the emission intensity was applied to all traded soybean and soybean cake, the approach equally distributes the LUC emissions across all importing countries irrespective of where the soybean is produced. Reduced time-frame approach. Annual deforestation rates are highly variable, so the period over which the rates of LUC are estimated can therefore have a significant influence on results. Since data from forestry inventories are only available from 1990, this assessment was based on the average rates of LUC over the period This not only coincides with a period of high rates of deforestation but also high soybean area expansion. In the reduced time frame approach, the LUC emissions are calculated based on the average rates of LUC over the period from , while maintaining the underlying assumptions in the study. 3.2 Results Effect of LUC approach on soybean LUC emission factor. Table C7 reports the LUC factors for soybean cake (kg CO 2 -eq per kg soybean cake) calculated using each of the approaches. The choice of method for estimating LUC EFs can strongly influence the emission intensity of livestock products and illustrates the complexity of analysing LUC processes. The PAS :2012 approach produces markedly different LUC emission factors due to the assumptions made regarding the land use category against which additional land for soybean production was gained and the relative share of this gain (Table C8). Unlike Brazil, Argentina has a higher EF using the default assumption (that expanded crop areas are derived from forest, grassland and perennial crops in equal proportion) than using the relative rates of change. The higher proportion of soybean cultivated on expanded areas in Argentina (76 percent) compared to Brazil (55 percent), combines with the default LUC assumptions, to give Argentina a higher soybean EF than Brazil under PAS :2012. The strength of the One-Soy approach is that it recognizes that global demand is a key driver of LUC. However, it penalizes those countries whose production is not directly associated with LUC and may not provide the right signals to producers and consumers of soybean. In the reduced time-frame approach, the emission intensity of soybean cake from Argentina and Brazil reduces by more than half. Average annual deforestation rates 135

14 Greenhouse gas emissions from pig and chicken supply chains Table C8. Proportion of expanded soybean area derived from each land use Land use category GLEAM approach PAS :2012 approach Brazil Argentina Brazil Argentina percentage Forest (33) 23 (33) Grassland (33) 0 (33) Shrubland (0) 0 (0) Annual cropland (0) 61 (0) Perennial cropland (33) 16 (33) Note: Figures in brackets are the PAS default land use transformations. Sources: Based on FAOSTAT (2012). appear to be close over the two periods and (1.76 and 1.98 million ha respectively, Figure C3), but the average annual rates of soybean expansion differ and they are higher for : between 1990 and 2006, the soybean area in Brazil increased by ha/year whereas the increase for the period was ha/year. The lower emission intensity for therefore results from the rate of deforestation relative to the rate of soybean expansion, not from the absolute change in deforestation rate. Effect of LUC approach on meat and egg emission intensity. In order to test the sensitivity of the results to different soybean LUC methods, the analysis of pigs and chickens in the UK and Viet Nam was rerun with the emission intensities calculated using the different LUC approaches. Results are given in Table C9. Within country effect. The effect of changing the soybean LUC approach on the emissions intensity of meat or eggs can vary between different combinations of system and species within the same country as a result of differences in the percentage of soybean in ration. For example, changing the soybean LUC approach results in a greater change in the emission intensity of UK broilers than layers (Table C10) as they have a greater percentage of soybean (and therefore soybean from Brazil) in their ration compared to UK layers (Table C12). Furthermore, the relative importance of feed emissions to total emissions intensity influences the proportionate increase in emissions intensity. For example, feed emissions make up a greater proportion of the broilers total emission intensity (as they have lower manure emissions and higher feed conversion ratios), so a 10 percent change in the feed emission intensity will lead to a greater increase in broilers than layers. Finally, soybean used in backyard systems is assumed to not be associated with LUC, so the emissions intensity within these systems therefor does not vary in response changing soybean LUC methods (Table C10 and C11). Between country effect. Differences in the total amount of soybean that is imported affect the emissions intensity of meat/eggs when using the One-Soy approach because with this method the EF for soybean varies depending on the percentage of the soybean that is imported, rather than the specific country that it is imported from. 136

15 Appendix C Changes in carbon stocks related to land use and land-use change Figure C3. Annual forest loss in Brazil Time period considered in this assessment hectares Table C9. Soybean land-use change emissions factors for the United Kingdom and Viet Nam in 2005 (kg CO 2 -eq/kg soybean DM) United Kingdom Viet Nam Soybean Soybean cake Soybean Soybean cake GLEAM approach (current study) PAS : One-Soy Reduced time-frame Source: Authors calculations. Table C10. Total emissions intensity for chickens (kg CO 2 -eq/kg meat/egg protein) Broilers: UK Layers: UK Backyard: Viet Nam GLEAM approach (current study) PAS : One-Soy Reduced time-frame Source: Authors calculations. Table C11. Total emissions intensity for pigs (kg CO 2 -eq/kg meat/egg protein) Industrial: UK Intermediate: Viet Nam Backyard: Viet Nam GLEAM approach (current study) PAS : One-Soy Reduced time-frame Source: Authors calculations. 137

16 Greenhouse gas emissions from pig and chicken supply chains Table C12. Proportions in the ration of soybean and soybean imported from Brazil and Argentina (percentage) Soybean in ration Soybean imported from Soybean in ration from Brazil Argentina Brazil Argentina Industrial pigs: UK Intermediate pigs: Viet Nam Backyard pigs: Viet Nam Broilers: UK Layers: UK Backyard chickens: Viet Nam Source: GLEAM, FAOSTAT (2012) and authors calculations. Differences in where the soybean is imported from, i.e. the amounts imported specifically from Brazil and Argentina, affect the emissions intensity of meat/eggs when using the GLEAM and PAS 2050 methods because with these methods the soybean EF varies depends on where the soybean is produced. Using GLEAM, Brazilian soybean has a higher EF than Argentinian (Table C7), so systems that have significant amounts of Brazilian soybean in their ration (such as UK broilers and industrial pigs see Table C12) will have higher LUC emissions (Table C10 and C11). However, using the PAS 2050 method Argentinian soybean has a higher EF than Brazilian, leading to a higher emission intensity for intermediate pig meat in Viet Nam than under the GLEAM method (see Table C11). In addition to the method used to calculate the soybean EF, the LUC emissions per kg of meat/eggs also depends on national differences in the amount of soybean in the ration and feed conversion ratios. Therefore, although two countries may import the same total amount of soybean, and the same amounts from Brazil and Argentina, the same species and system may still have quite different soybean LUC emissions. 4. COMPArison with OTHEr STudies The emission intensity for LUC per kg of soybean and soybean cake calculated in this study are compared to other studies in Table C13. The emission intensity used in this study is higher than some other studies, but within the overall range. The emission intensity of soybean is highly dependent on the method and assumptions used to calculate it (Flysjo et al., 2012). Variation arises from differences in: The calculation of C losses in soil and vegetation (above- and below-ground); The quantification of land-use transitions, i.e. how much of the LUC can be attributed to cropping; The ways in which the LUC emissions are allocated to specific crops. Emissions can be allocated in different ways, such as: (a) the crops grown in the country/ region where the LUC has occurred; (b) all expanding crops grown in the country/region where LUC has occurred; (c) all crops grown globally. These different allocation methods can lead to variations in the emissions per kg of crop; The time period over which emission are allocated. The estimates of LUC emissions presented in this report are still very preliminary and need to be interpreted with caution. This is an important area for improvement of GLEAM and it is planned that future developments of the model will include a more detailed and complete assessment of LUC emissions. 138

17 Appendix C Changes in carbon stocks related to land use and land-use change Table C13. Soybean land use-change emissions per unit of output and hectare Study Area covered by study Emissions *Converted/all soybean/all crops FAO (2010a) Argentina 1.04 kg CO 2 -eq/kg soybean all soybean FAO (2010a) Brazil 7.69 kg CO 2 -eq/kg soybean cake all soybean FAO (2010a) Brazil 8.54 kg CO 2 -eq/kg soybean cake all soybean FAO (2010a) Brazil kg CO 2 -eq/kg soybean cake converted FAO (2010a) Brazil kg CO 2 -eq/kg soybean converted Leip et al. (2010) grass>soybean South America 1.50 kg CO 2 -eq/kg soybean cake all soybean Cited in Flysjö et al. (2012) Leip et al. (2010) mix>soybean South America 3.10 kg CO 2 -eq/kg soybean cake all soybean Cited in Flysjö et al. (2012) Leip et al. (2010) forest>soybean South America kg CO 2 -eq/kg soybean cake all soybean Cited in Flysjö et al. (2012) Sonesson et al. (2009, p13) Brazil 1.50 kg CO 2 -eq/kg soybean all soybean ~0.6 of this is due to LUC Audsley et al. (2010, p.59) Audsley et al. (2010, p.59) Brazil 5.30 kg CO 2 -eq/kg soybean all soybean Argentina 1.60 kg CO 2 -eq/kg soybean all soybean Castanheira and Freire (2011) Low (Argentina) ~0.5 kg CO 2 -eq/kg soybean converted Castanheira and Freire (2011) High (Brazil) ~15 kg CO 2 -eq/kg soybean converted Nemecek et al. (2012) Brazil 1.47 kg CO 2 -eq/kg soybean all soybean Brazil, LUC, Ecoinvent v2.2 Nemecek et al. (2012) Brazil 5.21 kg CO 2 -eq/kg soybean all soybean Brazil, LUC, Ecoinvent v3.0 Reijnders & Huijbregts (2008) Brazil cerrado 1 to 2.7 kg CO 2 -eq/kg soybean converted Reijnders & Huijbregts (2008) Brazil forest 5 to 13.9 kg CO 2 -eq/kg soybean converted FAO (2010a) Brazil deforestation kg CO 2 -eq/ha converted FAO (2010a) Brazil deforestation kg CO 2 -eq/ha all soybean Audsley et al. (2009) All LUC 1.43 kg CO 2 -eq/ha allocates LUC to all crops globally Audsley et al. (2010, p.59) Brazil deforestation kg CO 2 -eq/ha converted Audsley et al. (2010, p.59) Brazil - grassland kg CO 2 -eq/ha converted Reijnders & Huijbregts (2008) Brazil forest 14 to 39 kg CO 2 -eq/ha converted Schmidt et al. (2011) All LUC 8.42 kg CO 2 -eq/ha allocates LUC to all crops globally *EF for (a) converted land; (b)average over all soybean grown in country/region; or (c) all crops grown globally. References Armentano, T.V. & Menges, E.S Patterns of change in the carbon balance of organic soil-wetlands of the temperate zone. Journal of Ecology, 74: Audsley, E., Brander, M., Chatterton, J., Murphy-Bokern, D., Webster, C. & Williams, A How low can we go? An assessment of greenhouse gas emissions from the UK food system and the scope to reduce them by Food Climate Research Network and WWF, UK. Audsley, E., Angus, A., Chatterton, J., Graves, A., Morris, J., Murphy-Bokern, D., Pearn, K., Sandars, D. & Williams, A Food, land and greenhouse gases. The effect of changes in UK food consumption on land requirements and greenhouse gas emissions. The Committee on Climate Change, London. 139

18 Greenhouse gas emissions from pig and chicken supply chains Bickel, U. & Dros, J. M The impacts of soybean cultivation on Brazilian ecosystems. Bonn, AIDEnvironment-WWF): p 33. (available at BSI PAS 2050:2008. Specification for the assessment of the life cycle greenhouse gas emissions of goods and services. British Standards Institution, UK. BSI PAS :2012. Assessment of life cycle greenhouse gas emissions from horticultural products Supplementary requirements for the cradle to gate stages of GHG assessments of horticultural products undertaken in accordance with PAS British Standards Institution, London. Brandao, A.S.P., Castro de Rezende, G. & Da Costa Marques, R.W Agricultural growth in the period , outburst in soybeans area and environmental impacts in Brazil. Discussion Paper No Instituto de Pesquisa Econômica Aplicada, Brasilia. Carvalho, G.O., Nepstad, D., McGrath, D., Vera Diaz. M. del C. & Barros, A.C Frontier expansion in the Amazon. Balancing development and sustainability. Environment Sci. Policy Sustainable Dev., 44: Castanheira, E. & Freire, F Life-cycle greenhouse gas assessment of soybeans. Paper presented at Life-Cycle Management Conference, 2011, Berlin. Cederberg, C., Persson, M.U., Neovius, K., Molander, S. & Clift, R Including carbon emissions from deforestation in the carbon footprint of Brazilian beef. Environmental Science & Technology, 45 (5): Coleman, K., Jenkinson, D.S., Crocker, G.J., Grace, P.R., Klir, J., Korschens, M., Poulton, P.R. & Richter, D.D Simulating trends in soil organic carbon in long-term experiments using RothC Geoderma 81: Conant, R.T. & Paustian, K Potential soil carbon sequestration in overgrazed grassland ecosystems. Global Biogeochemical Cycles, 16(4): Costa de Campos, B., Carneiro Amado, T. J., Bayer, C., Nicoloso, Rodrigo da Silveira, & Fiorin, J.E Carbon stock and its compartments in a subtropical oxisol under long-term tillage and crop rotation systems. Revista Brasileira de Ciência do solo, 35(3): Del Grosso, S.J., Parton, W.J., Mosier, A.R., Hartman, M.D., Brenner, L., Ojima, D.S. & Schimel, D.S Simulated interaction of carbon dynamics and nitrogen trace gas fluxes using the DAYCENT model. In Shaffer, J., Liwang Ma & S. Hansen, Schaffer, M. et al., eds. Modelling carbon and nitrogen dynamics for soil management, pp Boca Raton, CRC Press, Florida. Dieckow, J., Zendonadi dos Santos, N., Bayer, C., Molin, R., Favaretto, N. & Pauletti, V No-tillage crop rotations, C sequestration and aspects of C saturation in a subtropical Ferralsol th World Congress of Soil Science, Soil Solutions for a Changing World. Brisbane, Australia. FAO World livestock production systems: current status, issues and trends, by Seré, C. & Steinfeld, H. FAO Animal Production and Health Paper 127. FAO, Rome. FAO Greenhouse gas emissions from the dairy sector A life cycle assessment, by Gerber, P., Opio, C., Vellinga, T., Henderson, B. & Steinfeld, H. FAO, Rome. FAOSTAT FAO Statistical Database. Accessed Falloon, P. & Smith, P Simulating SOC changes in long-term experiments with RothC and CENTURY: model evaluation for a regional scale experiment. Soil Use and Management, 18: Fearnside, P.M Deforestation in Brazilian Amazonia: history, rates and consequences. Conservation Biology, 19(3):